This project would investigate a mechanistic understanding of children's knowledge about the organization of different categories and the role that knowledge plays in early nominal vocabulary growth. The proposed research concentrates on the mechanism of associative learning in the period between 13 and 36 months. It is during this time that object name learning becomes faster and more correct. The larger idea behind the proposal is that early noun learning teaches children about the perceptual similarities relevant to different kinds of categories and in so doing creates a self-accelerating developmental trajectory. Four kinds of studies are proposed: (1) studies of input to children--the correlations among linguistic devices, object properties, and category organizations; (2) studies of how children extend the nouns they know to new instances; (3) studies of children's novel noun extensions and their use of object properties, task context, and linguistic context to modulate their category formation; (4) studies of the self-accelerating and self-directing nature of early noun learning (including longitudinal studies, training studies, and studies of children who are late talkers). In addition, the mechanistic basis of the observed developmental trends will be pursued in simulation studies that are closely tied to the experimental studies. An associative learning device, a connectionist net, will be taught the nominal categories children learn. This tests whether the statistical regularities in the input are sufficient to create the observed learning biases. The modeling work includes modeling of individual developmental trajectories of children, including late talkers.